The SMR was almost three times higher for persons infected with Campylobacter in Sweden within the first months after disease onset. These risks diminished with time, and after a year or more have passed the SMR almost equaled that for the general population. For the imported cases of Campylobacter we observed lower SMRs throughout every time point in the study period.
Case fatality rate (CFR) is widely used and perhaps more intuitive than standardized mortality ratio when disease specific mortality is estimated. Up until November 1, 2004, 563 of the cases infected in Sweden, and 222 of the cases infected abroad had died. The CFR within 30 days was 0.19% (95% CI 0.13–0.27%) for those infected in Sweden, and 0.008% (95% CI 0.0008–0.03%) for those infected abroad. However, we do not favour the use of CFR because information on competing causes of mortality is missing, along with age effects.
The different SMRs depending on whether a case was infected at home or abroad must be analyzed in a context of how a case is diagnosed and reported. For gastrointestinal diseases there are problems with under-reporting and biases regarding which cases that eventually turns up in the national routine surveillance.
The problems with under-reporting affect every step in the chain: an ill person must seek health care, the doctor needs to take a stool sample, the laboratory must be capable of accurately detecting the pathogen, and if the laboratory find the pathogen the doctor needs to send in a notification. It is perhaps in the first of these steps that national surveillance misses most of the cases. Studies in both United Kingdom and United States have tried to measure the proportion reported to national surveillance from those who have symptoms of infectious intestinal disease in the population. For an agent like Campylobacter that typically causes non-bloody diarrhea a multiplier in the range of 7.6–38 has been estimated [10, 11].
All persons who get infected with Campylobacter do not have the same chance of becoming notified as a Campylobacter case. There are three different groups that are more likely than others to be reported [12]. Firstly, people who have a severe disease or profound symptoms are more likely to see a doctor, compared to others. Secondly, persons with a history of recent travel prior to the onset of symptoms are also more likely to seek health care and become diagnosed and reported. Thirdly, persons with pre-existing illnesses of certain magnitude may be over-represented in the national surveillance statistics.
These biases in disease reporting will have an impact on our findings and can help us in the interpretation of the results. The general under-reporting will affect milder cases in particular. Therefore, the persons in our cohort infected in Sweden were most likely the more severe cases in the population. On the other hand, travellers who develop symptoms of gastro-enteritis are more likely on the average to have a physical exam when they have returned home and they do not necessary need to have a severe disease. The finding of a standardized mortality ratio equal to 1.0 after one year had passed after the infection suggest that the mortality of this cohort infected domestically is comparable to the standard Swedish population after one year. This does not necessarily imply that the cohort is comparable to the Swedish population also at the earlier time points. The risk profile may have changed due to frailty. The statistical term frailty implies that all individuals do not have the same risk profile and frailer individuals may succumb earlier and therefore the risk profile of a cohort changes over time [13].
It is difficult to find information on Campylobacter associated mortality. Some have estimated the case fatality rate to be in range of 1–3/10000 cases [14]. Others that have used an approach more like ours found that 1.2% (16180 cases and 190 deaths) where deceased within one year from diagnosis [15]. Unfortunately, no study has compared mortality risks among travellers and non-travellers for campylobacteriosis.
Our findings suggest that the individuals reported into the surveillance system are a highly selected. Individuals with a domestically contacted infection may have a different, more severe risk profile, compared to individuals in the cohort with an imported infection. The reason for these differences is not characteristics of the infection itself but reflects differences in how these individuals are selected into the surveillance system. Any registry-based estimation of mortality among cases of campylobacteriosis must therefore take this into account and stratify cases according to the place of infection.